Newsletter: Agent outcome = "Success"—Or Is It?
What happens when AI can do anything (like generating 17 unnecessary PowerPoints)?
Brixo latest
Engineers didn’t sign up to write reports. But that’s where a lot of AI teams end up.
The business wants to know:
“How’s the agent performing?”
Engineering pulls up traces, checks logs, scans retries, and replies with:
“Success rate: 94%.”
From an observability standpoint — it looks great!
But what that data doesn’t show is that 40% of those “successful” runs came from frustrated users.
One example we saw: a support agent completed the task, but the customer was so annoyed they never returned.
Engineering saw a “success!”
The business saw a churned customer.
That’s the disconnect.
Observability tools are great — for engineers. They track latency, errors, and retries. But they don’t explain what the agent means for the business.
Engineers get stuck translating telemetry into something everyone else can understand.
That’s why we built Brixo — a translation layer between engineering and the business.
It turns traces, retries, and token logs into clear answers about performance, cost, and customer impact.
So engineering can keep building — and the business can finally see what’s really happening.
What stood out
Matt Turck, a VC @FirstMarkCap, is spot on.
What we read/listened/watched
A16z released a top 50 list of AI products that startups spend on.
As someone that is pretty in-tune with AI products, I’ll admit I had never heard of 3 of the top 10!
If you haven’t heard it yet, Stripe cofounder, John Collison, has a great podcast. Here he interviews Marc Andreesen.
Ethan Mollick shows AI doing genuinely impressive things (like replicating academic research in minutes) while warning us with a perfect example: he had Claude generate 17 PowerPoint versions of the same memo—because it could, not because anyone needed them.
Mira Murati’s Stealth AI Lab Launches Its First Product
Mira Murati’s new startup Thinking Machines Lab just launched Tinker, a tool that lets anyone fine-tune cutting-edge AI models without needing massive computing resources or technical expertise.
Reminder: Thinking Machines raised a $2B seed round. Yes, you read that right.
Final Brick
Every wondered what the largest seed investment valuations ever are?
The top five largest seed investment valuations ever recorded are dominated by recent AI startups, with sizes and valuations that far outpace historical norms.
Thinking Machines is 6.7X larger than second place.
The five biggest seed rounds and their reported valuations as of October 2025:
